Pachyderm Examples

Pachyderm Examples is a curated list of examples that use Pachyderm to accomplish various tasks.

Getting Started

• Intro to Pachyderm Tutorial - A notebook introduction to Pachyderm, using the pachctl command line utility to illustrate the basics of Pachyderm data repositories and pipelines
• Boston Housing Prices - A machine learning pipeline to train a regression model on the Boston Housing Dataset to predict the value of homes.
• Boston Housing Prices (Intermediate) - Extends the original Boston Housing Prices example to show a multi-pipeline DAG and data rollbacks.
• Market Sentiment - Train and deploy a fully automated financial market sentiment BERT model. As data is manually labeled, the model will automatically retrain and deploy.
• Object Detection - Train an object detector on the COCO128 dataset with Lightning Flash, modify predictions with Label Studio, and version everything in Pachyderm.

Notebooks

• JupyterLab Pachyderm Mount Extension - A notebook showing how to use the JupyterLab Pachyderm Mount Extension to mount Pachyderm data repositories into your Notebook environment.
• Jsonnet Pipeline Specs - A notebook introducing and showing how use Jsonnet Pipeline Specs to templatize common pipelines.

Data Labeling

• Label Studio Integration - Incorporate data versioning into any labeling project with Label Studio and Pachyderm.
• Superb AI Integration - Version labeled image datasets created in Superb AI Suite using a cron pipeline.
• Toloka Integration - Uses Pachyderm to create crowdsourced annotation jobs for news headlines in Toloka, aggregate the labeled data, and train a model.

Data Warehouse

• Churn Prediction with Snowflake - Create a churn analysis model for a music streaming service with Pachyderm and Snowflake using the Data Warehouse integration.

Machine Learning

• Boston Housing Prices (Intermediate) - Extends the original Boston Housing Prices example to show a multi-pipeline DAG and data rollbacks.
• Breast Cancer Detection - A breast cancer detection system based on radiology scans scaled and visualized using Pachyderm.
• Market Sentiment - Train and deploy a fully automated financial market sentiment BERT model. As data is manually labeled, the model will automatically retrain and deploy.
• Apache Spark - MLflow Integration - End-to-end example demostrating the full ML training process of a fraud detection model with Spark, MLlib, MLflow, and Pachyderm.

ML Experiment Tracking

• Weights and Biases - Log pipelines running in Pachyderm to Weights and Biases.
• ClearML Integration - Log Pachyderm experiments to ClearML's experiment montioring platform, using Pachyderm Secrets.

Model Deployment

• Pachyderm - Seldon - Community example showing monitoring and provenance for machine learning models with Pachyderm and Seldon.
• Seldon (Market Sentiment) - Deploy the model created in the Market Sentiment example with Seldon Deploy.